multiple purpose spin transfer torque operated …facta.junis.ni.ac.rs/eae/fu2k133/eae130306.pdf2.1....
TRANSCRIPT
FACTA UNIVERSITATIS Series: Electronics and Energetics Vol. 26, No 3, December 2013, pp. 227 - 238 DOI: 10.2298/FUEE1303227W
MULTIPLE PURPOSE SPIN TRANSFER TORQUE
OPERATED DEVICES
Thomas Windbacher, Hiwa Mahmoudi, Alexander Makarov,
Viktor Sverdlov, Siegfried Selberherr
Institute for Microelectronics, TU Wien, Gußhausstraße 27-29/E360, 1040 Wien
Abstract. We summarize our recent work on a non-volatile logic building block required
for energy-efficient information processing systems. A sequential logic device, in
particular, an alternative non-volatile magnetic flip-flop has been introduced. Its
properties are investigated and its extension to a very dense shift register is demonstrated.
We show that the flip-flop structure inherently exhibits oscillations and discuss its spin
torque nano-oscillator properties.
Key words: spin transfer torque, non-volatile logic, flip-flop, latch, shift register, nano-
oscillator
1. INTRODUCTION
Due to the rising efforts to scale CMOS devices down with every technology node,
according to the ITRS [1], new technological processes incorporating global and local
strain, new materials for high-k/metal gates, and new fin-FET and three-dimensional tri-
gate device structures must be introduced. At the same time power consumption and in-
terconnect delay start to become an important performance bottleneck [2]. One possible
way to reduce the power consumption of CMOS circuits is to shutdown latent circuit
parts completely and to power them up only when they are needed. This transition from
normally "ON" to normally "OFF" circuits requires the successive reevaluation of all
computational building blocks. The use of non-volatile elements which do not need
power to keep their state is paramount for the success of this new type of information
processing. At this point the exploitation of spin as degree of freedom comes in handy,
due to its non-volatility, high endurance, and fast operation [3]. Furthermore, it allows
not only to store information efficiently, but also to introduce new approaches to how the
actual information is processed and the information transport is handled [4], [5]. Thus, it
also allows to look for alternatives to the von Neumann architecture, where successively
information is moved between the memory unit and the processing unit through a - now-
adays performance limiting - common bus.
Received December 12, 2013
Corresponding author: Thomas Windbacher Institute for Microelectronics, TU Wien, Gußhausstraße 27-29/E360, 1040 Wien, http://www.iue.tuwien.ac.at
(e-mail: [email protected])
228 T. WINDBACHER, H. MAHMOUDI, A. MAKAROV, V. SVERDLOV, S. SELBERHERR
The most obvious potential application of non-volatile spintronic devices as CMOS
alternative is dynamic and static random access memory. The discovery of the giant
magneto resistance (GMR) [6], [7] and the tunnel magneto resistance (TMR) [8] effects
further boosted interest in magnetic devices, which results in improving their perfor-
mance and bringing them to commercial applications [9], [10]. At its earlier stage this
research led to the development of magnetic memory arrays [9], [11] for random access
memory applications: MRAM. Initially, their logic states had to be written by a magnetic
field, requiring an extra wire for field generation. Unfortunately, this caused high writing
energies and unfavorable scaling behavior and thus was prohibitive for (very) large scale
integration (VLSI). This problem was resolved by first theoretically predicting and later
experimentally justifying the spin transfer torque effect [12], [13]. The spin transfer
torque (STT) effect allows a purely electric control of the magnetization orientation and
thus leaves the writing wire generating the magnetic field in MRAM superfluous. How-
ever, the current value needed to invert the magnetization of the free layer in magnetic
tunnel junction (MTJ) is still rather high, and its reduction is the main goal of research to
make this technology competitive with CMOS. Utilization of MTJs with perpendicular
magnetic anisotropy allows reducing the required writing power to a level, where it can
compete with CMOS SRAM cache [14], [15], [16], [17]. Several products based on spin
transfer torque magnetic random access memory (STT-MRAM) technology have been
recently introduced [18], [19], [20], [21].
However, besides the apparent MTJ applications for memory cells, new spintronic
devices open up additional opportunities for replacing CMOS based counterparts [3].
Below we introduce a non-volatile magnetic flip-flop as an information processing unit.
2. NON-VOLATILE MAGNETIC FLIP-FLOP
As already mentioned, the ever increasing demand in fast and cheap memory as well
as in electronics in general has driven the scaling efforts in CMOS since its very begin-
nings. Today, pushing the limits of integration density is still a major concern, but power
efficient computing gains more and more momentum. In addition to memory and combi-
natorial logic mostly employed in modern integrated circuits, sequential logic attracts
close attention. Sequential logic can be based on memory elements for information pro-
cessing. Its present output states not only depend on its present inputs, but also on its in-
put history [22]. Sequential logic devices are fundamental building blocks of digital elec-
tronics, and flip-flops as well as latches belong to this group of logic. The transition from
normally "On" to normally "Off" logic requires a redesign of these devices as well.
Even though already several circuit designs have been proposed to meet this goal,
e.g., Zhao et al. [23], these circuits are commonly CMOS-MTJ hybrids. The non-volatil-
ity is introduced by the MTJs, however, the MTJs act merely as ancillary devices by only
holding the information without any further functionality. The actual logic operations are
carried out in the CMOS elements. The introduction of non-volatility through MTJs adds
complexity because of the need for converting the information between resistance states
of the MTJs and voltage or current signals required for CMOS operation every time data
is written, read, or transferred.
Multiple Purpose Spin Transfer Torque Operated Device 229
Therefore, we propose a novel non-volatile magnetic flip-flop which shifts the actual
logic operation and thus the information processing from the electric signal CMOS
domain to the magnetic domain [24]. This allows creating denser and simpler layouts as
well as harvesting the beneficial features related to spintronics.
Fig. 1 Proposed non-volatile magnetic flip-flop
2.1. Structure and working principle
The non-volatile magnetic flip-flop consists of three fixed anti-ferromagnetically cou-
pled polarizer stacks (see Fig. 1). Two polarizer stacks are for input (A, B) and one po-
larizer stack is for readout (Q). They exhibit a fixed perpendicular magnetization orienta-
tion (parallel to the z-axis) and due to their anti-ferromagnetic configuration negligible
stray fields are assumed. The polarizer stacks are positioned on top of tunnel barriers (e.g.
Cu, Al2O3, and MgO) and share a common free layer. The common free layer exhibits a
perpendicular anisotropy with a constant uni-axial anisotropy K1 (parallel to the z-axis). It
is further assumed that the ratio of the common free layer is 4 to 1 with the size of the
short side of the layer ranging from 10 nm to 30 nm. The long side of the free layer is
along the y-axis, while the short side lies parallel to the x-axis.
Furthermore, a positive electric current is defined as flowing from the contacts (A, B,
Q) through the free layer (negative z-direction), while at the same time electrons flow in
the opposite direction (positive z-direction).The logic information is stored via the mag-
netic orientation of the free layer, and the polarity of the input pulses is mapped to logic
"0" and "1", respectively.
A current pulse applied to one of the input stacks (A or B) exerts a spin transfer
torque on the magnetization in the corresponding free layer region of the input stack and
generates a precessional magnetization motion. This precessional motion couples through
the magnetic exchange to the rest of the free layer and creates a spin wave which travels
through the common free layer heading to its opposite end, where it is reflected and
moves back, gets reflected, and pushed again and so on [25]. During this kind of oscillat-
ing spin wave motion, the localized magnetic moments in the common free layer are ex-
cited and their precessional motions build up, until the magnetization eventually passes
the energy barrier between the two stable states and relaxes into the other stable state fast.
If now, instead of one, two synchronous current pulses are applied to the input stacks A
and B, two moving spin waves are generated. Depending on their polarity they either try
to enforce the switching of the free layer magnetization or hold it in its current position.
There are four possible input combinations. In the case of two positive currents both
generated torques align the magnetization parallel to the z-axis. Therefore, the magneti-
zation orientation stays unchanged. If both input currents are negative, both generated
230 T. WINDBACHER, H. MAHMOUDI, A. MAKAROV, V. SVERDLOV, S. SELBERHERR
torques try to flip the magnetization orientation. In this case both spin waves add up
leading to a faster switching compared to a single input current. Finally, in the case of
opposing input current polarities one torque always tries to hold the current magnetiza-
tion state, while the other strives to flip it. Since these two contributions compensate each
other, the magnetization direction of the free layer remains the same.
Thus, two sufficiently high and long pulses with identical polarity either write logic
"0" or "1" into the common free layer, while two pulses with opposing polarities cancel
each other and the initial magnetization state is held representing a sequential logic func-
tion needed for flip-flop logic. If one of the inputs is inverted, the resulting logic sets or
resets the held state, when the input signals are opposing, and holds its state for identical
input polarities. This corresponds to RS flip-flop logic, but without forbidden input com-
binations as compare to its transistor counterpart [22].
2.2. Device characteristics
In order to test our proposed device a series of micromagnetic simulations was carried
out [26], [27]. Three different free layer sizes with length to width ratio of four to one
were studied. It was further assumed that the common free layer featured layer widths of
10nm, 20nm, and 30nm, a layer thickness of 3nm, a perpendicular uni-axial anisotropy
K1=105 J/m
3 along the z-axis, and that the spin barrier connecting the polarizer stacks to
the free layer is made out of copper (cf. Fig. 1). The saturation magnetization of the
common free layer was considered as MS=4×105A/m, an exchange constant of
Aex=2×10-11
J/m, and a spin current polarization P of 0.3. As mentioned in Section 2.1, the
polarizer stacks consist of anti-ferromagnetically coupled layers. Therefore, stray fields
are neglected.
The magnetization dynamics is described by the Landau-Lifshitz-Gilbert equation
[28], [29]:
T
dt
mdmHm
dt
mdeff
, (1)
where ⃗⃗ denotes the reduced magnetization, γ=2.211×105m/As the Gilbert gyromagnetic
ratio, α = 0.01 the dimensionless damping constant, ⃗⃗ the effective field in A/m. The
last term in (1) describes the spin transfer torque ⃗ with the following spin transfer torque
model [30]:
)´()1()1( 22
2
0
pmmpmpm
P
Ml
J
eT
S
. (2)
denotes the Planck constant, µ0 the permittivity of vacuum, e the electron charge, J the
applied current density, l the free layer thickness, MS the magnetization saturation, P the
polarization, the unit polarization direction of the polarized current, and Λ=2 a fitting
parameter handling non-idealities. Furthermore, ⃗⃗ is calculated from the functional
derivative of the free energy density functional containing uni-axial anisotropy, ex-
change, and demagnetization contributions [31].
Multiple Purpose Spin Transfer Torque Operated Device 231
Fig. 2 Switching times for identical input pulses as a function of applied current density
Fig. 3 Switching times for opposing input pulses as a function of applied current density. m
denotes a negative polarity current pulse, while p denotes a positive polarity current
pulse. For instance pmp (A, B, free layer) means a positive current pulse applied at
contact A, a negative current pulse applied at B, and a magnetization orientation of
the free layer along z-axis
The data points in Fig. 2 and Fig. 3 are averaged over 100 samples. The error bars de-
pict an error range of ±3σ. Fig. 2 shows the switching times for the set and reset operation
as a function of applied current densities and different layer sizes. The corresponding
switching times are defined as the time it takes from the onset of the current pulse until
80% of the final magnetization state is reached. One can see that at low current densities
it takes relatively long to build up the required precessional motion for flipping the layer
magnetization, while with increasing current densities the switching times decreases fast.
The difference in the threshold current densities for different geometries can be explained
as follows. Reducing the layer width reduces the demagnetization factor along z but at
the same time the demagnetization factors along x and y start to increase [32]. This leads
to a smaller depolarization field along z and a bigger net barrier parallel to the z-axis,
since the depolarization field opposes the anisotropy field. Therefore, the smaller the
layer width the higher is the switching barrier and the required current to pass the barrier.
For a pulse length of 20ns switching starts at 4×1010
A/m2 (16±3.5ns), 5×10
10A/m
2
232 T. WINDBACHER, H. MAHMOUDI, A. MAKAROV, V. SVERDLOV, S. SELBERHERR
(13±3ns), and 6×1010
A/m2 (17.5±4.3ns) for 30nm, 20nm, and 10nm width, respectively. It is
worth to mention that, the switching speed becomes almost size independent around
≈1011
A/m2 until ≈5×10
12A/m
2 and the found switching times are in the order of ≈20ps to
≈100ps.
In contrast to the set and reset operation, the hold operation must not lead to a change
in the magnetization state. This is the case at opposing input pulses (see Fig. 3). For a
reliable flip-flop operation the magnetization ⃗⃗ must relax back to its initial orientation,
when the currents are turned off. Starting with an initial magnetization state along the
positive z-axis (as shown in Fig. 1), the total average magnetization projection should not
become negative during its precessional motion. If the magnetization overcomes the state
separating energy barrier and its projection on the z-axis becomes negative ( ), the
final magnetization state starts to depend on the pulse length. In other words, this is the
point, when the held information is lost.
The devices behave as required for current densities below 2×1012
A/m2 for 10nm
width, 7×1011
A/m2 for 20nm width, and 5×10
11A/m
2 for 30nm width (cf. Fig. 3). Close to
the respective current densities the spin wave generated at A (B) is not sufficiently
damped anymore and oscillatory motions start to build up. In fact, above the shown limits
the oscillations can become so strong that they cross the xy-plane and the initial
magnetization orientation is lost. Due to the geometrical confinement the oscillation
frequencies of the excited spin waves as well as the onset current densities for their
excitation depend on for the shared free layer sizes.
Already at this point the proposed non-volatile magnetic flip-flop is very attractive. In
addition, it requires less space than its CMOS counterparts, e.g., eight or twelve
transistors for a classical simple or clocked RS flip-flop, correspondingly [22], and it is
also more space efficient than hybrid MTJ/CMOS circuits, for example, the flip-flop
proposed in [23], which requires seven transistors and two magnetic tunnel junction
memory elements.
Instead the non-volatile magnetic flip-flop needs only three magnetic stacks (spin
valve and/or MTJ) sharing a common free layer. Even more, there are no forbidden input
combinations ( ̅ ) making the circuit and logic design easier. Due
to its back end of line compatibility and stack friendly topology (see Fig. 4), it is also
attractive for large scale integration.
2.3. Shift-register
In [33] we proposed an application of the flip-flop for a non-volatile shift register.
The shift register features an extremely dense layout and allows further reduction of the
required space (cf. Fig. 4).
The novel shift register topology takes advantage of the peculiarities of the non-
volatile flip-flop. Arranging two rows of flip-flops in two different levels along a line (as
shown in Fig. 3) or in a cross-like structure [33] and applying two time shifted clock
signals (Clk1 and Clk2, see Fig. 5 and Fig. 6), the information contained in one flip-flop
can be successively passed to the next subsequent flip-flop via the STT effect.
The shift register works as follows. Each flip-flop is alternately operated in a writing
and a reading mode. In the writing mode the inputs Ai and Bi are active, while Qi is
inactive. In this mode one input is fed with the information/signal while the other is fed
by one of the two clock signals (depending on its level, cf. Fig. 5). In the reading mode Ai
Multiple Purpose Spin Transfer Torque Operated Device 233
and Bi are inactive and Qi is active. A save decoupling between the two operation modes
is achieved by a time shift between the two clock signals (see Fig. 6).
In order to avoid the unintentional flipping of the common free layer (also known as
read disturbance), the pulse time mismatch between the reading and writing mode is
used. While in reading mode only the part of the common free layer under the Qi region
experiences a spin torque, in writing mode two regions Ai and Bi exert a spin torque on
the common free layer. Therefore, it takes about twice as long to change the magnetiza-
tion orientation in reading mode than in writing mode provided the currents are the same.
Let us illustrate the working principal with an example (see Fig. 4, Fig. 5, and Fig. 6).
Applying an input pulse to A1 and a first clock signal Clk1 to B1 to the first flip-flop
(Fig. 4, left) will set or reset the flip-flop, if both pulses exhibit the same polarity. When
the input signal and the clock signal Clk1 exhibit the same torque on the shared free layer,
information is written into the first free layer and can be accessed via Q1. At the next step
clock signal Clk1 turns inactive and Clk2 becomes active (cf. Fig. 6). This time the first
flip-flop is in reading mode and a current passes through the first free layer and the spin
barrier separating the two neighboring free layers before it enters to the second free layer
and exerts a spin torque on the second free layer. At the same time the clock signal Clk2
creates spin torque in the B2 region and the interaction between the two spin torque
sources either aids or suppresses the switching of its magnetization state. Afterwards the
clock signal Clk2 is switched off again and the clock signal Clk1 is turned on. Hence, the
second flip-flop is now in reading mode and information is passed via the STT effect to
the third free layer and so on. It is necessary to mention that the voltage or current signals
passing the information at Qi must be synchronized with respect to the corresponding
clock signals and that due to the additional degree of freedom introduced by the variable
shared free layer magnetization, Qi must be fed with one type of polarity only. This is
necessary to compensate the changing free layer magnetization in region Qi. The clock
signals Clk1 and Clk2 require a positive and negative pulse within their signal period to
guarantee that the information stored in one shared free layer will be successfully passed
into the next subsequent flip-flop, since only by two identical input pulses the flip-flop is
written.
Fig. 4 Proposed non-volatile magnetic shift register
Fig. 5 Circuit diagram of non-volatile magnetic shift register
234 T. WINDBACHER, H. MAHMOUDI, A. MAKAROV, V. SVERDLOV, S. SELBERHERR
Fig. 6 Signal timing example for the non-volatile magnetic shift register showing
the successive propagation of the input signal through register
2.4. Intrinsic oscillator
Despite its use as a sequential logic building block the flip-flop structure also features
large and stable oscillations (see Fig. 3 and Fig. 7). This is bad for logic applications, but
quite appealing for exploitation as a nano-scale oscillator. Oscillators constitute a very
basic building block and are ubiquitous in modern electronics. They are needed in meas-
urement, navigation, communication systems, and more. Their periodic output signals are
used manifold, for example, for clocking digital circuits, generating electromagnetic
waves, in high speed digital systems, and many more. Spin torque nano-scale oscillators
are so attractive because of their cost effective on-chip integration as microwave oscilla-
tors, their nano-scale size, frequency tunability, broad temperature operation range, and
CMOS technology compatibility [34], [35], [36], [37], [38]. Even though substantial pro-
gress has been achieved and they can now operate without an external magnetic field,
their still rather low output power limits their practical implementation [39], [40].
As discussed in Section 2.2, we observe oscillations in the common free layer, if cur-
rent densities with opposing polarities above the respective thresholds are applied (shown
in Fig. 3) [33], [41]. The excited precessional motions form large and stable orbits and do
not require an external magnetic field (see Fig. 7).
Fig. 7 Plot of the averaged and normalized magnetization
of the common free layer for 10nm, 20nm, and 30nm, respectively
Multiple Purpose Spin Transfer Torque Operated Device 235
Fig. 8 Frequency as a function of applied current density and free layer size
The occurring precessional motions are a superposition of two movements - an in-
plane precession (fx,z) and an out-of-plane oscillation (fz). Plotting the found oscillation
frequencies as a function of applied current density and free layer size shows that for a
fixed current density the frequency fz for the out-of-plane component oscillations (filled
symbols) is two times the corresponding in-plane precession fx,y frequency (open sym-
bols). It also can be observed that for different shared free layer sizes the precessional
frequencies change (see Fig. 8). For the 30nm×120nm structure the stable precessional
motion starts at ≈250 MHz (≈5×1011
A/m2) and increases up to ≈450 MHz (1×10
12A/m
2),
while for 20nm×80nm the precessions start at ≈1 GHz (1.×1012
A/m2) and increases to
≈ 1.17 GHz (1.25×1012
A/m2). For the 10nm×40nm structure they range from ≈3.22 GHz
(1.6×1012
A/m2) to ≈ 3 GHz (2.05×10
12A/m
2).
Further analysis shows that in the case of a single pulse at one input polarizer the
magnetization of the free layer is inverted, even though the generated spin torque acts
only on the free layer below the corresponding polarizer stack. Therefore, in this setup
two input pulses are required for the generation of oscillation. This hints already to the
explanation how the oscillations are excited. For example, if a positive current pulse is
applied through region A (shown in Fig. 1), the generated polarized current will exert a
spin torque on the local magnetic momenta and excite local precessional motions, which
try to flip the free layer magnetization. Through the exchange coupling these precessional
motions start to move out of the region where they are generated and travel through the
whole free layer (also known as spin wave). When the spin wave reaches the opposite
end of the free layer (B), it experiences a second torque which tries to preserve the mag-
netization orientation of the free layer (negative current). The wave gets reflected, heads
back to region A where it is pushed out again, and so on. The result is a stable preces-
sional motion in the whole free layer. This explanation is consistent with the time and
position resolved magnetization simulations we conducted and also explains the depend-
ence of the frequency on the free layer size [38]. While other groups reported a localiza-
tion of the oscillations below the spin injection site, in our device the full layer contrib-
utes to the precessions [36], [37], and [38]. We think this difference stems from the rather
strong stray field in the setups used in [36], [37], and [38], while in our case the stray
field is negligible.
However, in order to extract power from the oscillator the projection of the preces-
sional motion on the pinned layer magnetization orientation should be as big as possible.
236 T. WINDBACHER, H. MAHMOUDI, A. MAKAROV, V. SVERDLOV, S. SELBERHERR
Unfortunately, the perpendicular polarizers (see Fig. 1) lead to an extraction of the much
weaker vertical oscillations fz (see Fig. 7). A possible way to circumvent this obstacle is
to change the orientation of the output polarizer Q to an in-plane configuration and ex-
tract the signal from there. Thus, the much bigger component can be used. But
this also comes at a price; since the power extracted from Q is proportional to the power
put through Q, it must be sufficiently small to not disturb the oscillations.
Therefore, one might instead of changing Q to an in-plane orientation, modify A and
B to in-plane configurations [39]. This case was tested in [41], where it was shown that
the free layer magnetization also preforms large in-plane precessions. The observed sta-
ble oscillations exist even when the current is driven through a single stack. The current
through the second stack allows tuning the frequency. If the current through the second
stack exhibits the opposite polarity, the frequency of the oscillations is almost doubled.
Because of the in-plane orientations of the polarizer stacks, the large component
is extracted and the available microwave power is increased. Because the whole free
layer exhibits the oscillations when the excitation sources are at its opposite ends, wire-
less power extraction and further distribution is an interesting option for this structure.
3. OUTLOOK AND CONCLUSION
Even though great progress has been made over the last years, there are still chal-
lenges to be resolved to bring spintronics into mainstream electronics. One reason is the
maturity and excellence of CMOS technology itself. While the switching power of STT
based memory has become comparable with that of CMOS based SRAM and DRAM
[10], [14], [15], [16], and [17] its integration density is currently not as good as in state of
the art CMOS solutions. Spintronics for logic applications is under extensive develop-
ment. One of the show stoppers is the rather high energy required for switching: switches
based on the STT effect require still about 100 times more switching power than a CMOS
transistor [3]. Thus, the real challenge to bring the spin-driven devices close to commer-
cial applications is to reduce the required switching power for memory and logic. For
STT based nano-oscillators an increase of the output power is a pressing issue. A prom-
ising path to reach this goal is to migrate from current controlled switching to voltage
controlled switching. There are many promising results related to the voltage control of
magnetic properties. For instance, Wang et al. [42] have shown electric-field-assisted
reversible switching in CoFeB/MgO/CoFeB magnetic tunnel junctions with interfacial
perpendicular magnetic anisotropy, where the coercivity, the magnetic configuration and
the tunneling magnetoresistance can be manipulated by voltage pulses associated with
much smaller current densities (≈1.2×106A/m
2). Despite the critical switching power re-
duction also other essential features like fast switching in the below nanosecond regime
[43] and voltage control of the exchange bias at room temperature have been demon-
strated [44]. It also has been shown by Nozaki et al. [45] that electric-field-induced ferro-
magnetic resonance excitation by means of voltage control over the magnetic anisotropy
in a few monolayers of FeCo at room temperature is feasible. So we conclude that there
is still plenty of space left to explore and to further develop spintronics and its application
in large scale applications.
Acknowledgement: This work is supported by the European Research Council through the grant
#247056 MOSILSPIN.
Multiple Purpose Spin Transfer Torque Operated Device 237
REFERENCES
[1] International Technology Roadmap for Semiconductors, Chapter PIDS, [Online], Available:
http://www.itrs.net/Links/2012ITRS/Home2012.htm . [2] N. S. Kim, T. Austin, D. Baauw, T. Mudge, K. Flautner, J. S. Hu, M. J. Irwin, M. Kandemir, and V. Narayanan,
"Leakage Current: Moore's Law Meets the Static Power," Computer, vol. 36, no. 12, pp. 68–75, 2003.
[3] D. E. Nikonov and I. A. Young, "Overview of Beyond-CMOS Devices and a Uniform Methodology for Their Benchmarking," In Proceedings of the IEEE, vol. 101, no. 12, pp. 2498–2533, 2013.
[4] H. Mahmoudi, T. Windbacher, V. Sverdlov, and S. Selberherr, "Reliability-Based Optimization of Spin-
Transfer Torque Magnetic Tunnel Junction Implication Logic Gates," Adv. Mat. Res., vol. 854, pp. 89–95, 2014.
[5] H. Mahmoudi, T. Windbacher, V. Sverdlov, and S. Selberherr, "Implication Logic Gates Using Spin-
Transfer-Torque-Operated Magnetic Tunnel Junctions for Intrinsic Logic-In-Memory," Solid State
Electron., vol. 84, pp. 191–197, 2013. [6] M. N. Baibich, J. M. Broto, A. Fert, F. Nguyen Van Dau, F. Petroff, P. Etienne, G. Creuzet, A.
Friederich, and J. Chazelas, "Giant Magnetoresistance of (001)Fe/(001)Cr Magnetic Superlattices," Phys. Rev. Lett., vol. 61, no. 21, pp. 2472–2475, 1988.
[7] G. Binasch, P. Grünberg, F. Saurenbach, and W. Zinn, "Enhanced Magnetoresistance in Layered Magnetic
Structures with Antiferromagnetic Interlayer Exchange," Phys. Rev. B, vol. 39, iss. 7, pp. 4828–4830, 1989. [8] M. Julliere, "Tunneling Between Ferromagnetic Films," Phys. Lett. A, vol. 54, pp. 225–226, 1975.
[9] C. Chappert, A. Fert, and F. Nguyen Van Dau, "The Emergence of Spin Electronics in Data Storage,"
Nature Materials, vol. 6, pp. 813–823, 2007. [10] S. Ikeda, J. Hayakawa, Y. Ashizawa, Y. M. Lee, K. Miura, H. Hasegawa, M. Tsunoda, F. Matsukura, and H.
Ohno, "Tunnel Magnetoresistance of 604% at 300K by Suppression of Ta Diffusion in CoFeB/MgO/CoFeB
Pseudo-Spin-Valves Annealed at High Temperature," Appl. Phys. Lett., vol. 93, pp. 082508, 2008. [11] M. Hosomi, H. Yamagishi, T. Yamamoto, K. Bessho, Y. Higo, K. Yamane, H. Yamada, M. Shoji, H.
Hachino, C. Fukumoto, H. Nagao, and H. Kano, "A Novel Nonvolatile Memory with Spin Torque
Transfer Magnetization Switching: Spin-Ram," IEEE International Electron Devices Meeting (IEDM), pp. 459–462, 2005.
[12] L. Berger, "Emission of Spin Waves by a Magnetic Multilayer Traversed by a Current," Phys. Rev. B,
vol. 54, iss. 13, pp. 9353–9358, 1996. [13] J. C. Slonczewski, "Current-Driven Excitation of Magnetic Multilayers," J. Magn. Magn. Mater.,
vol. 159, iss. 1–2, pp. L1–L7, 1996.
[14] S. Ikeda, K. Miura, H. Yamamoto, K. Mizunuma, H. D. Gan, M. Endo, S. Kanai, J. Hayakawa, F. Matsukura, and H. Ohno, "A Perpendicular-Anisotropy CoFeB–MgO Magnetic Tunnel Junction," Nature Materials, vol. 9,
pp. 721–724, 2010.
[15] K. Abe, H. Noguchi, E. Kitagawa, N. Shimomura, J. Ito, and S. Fujita, "Novel Hybrid DRAM/MRAM Design for Reducing Power of High Performance Mobile CPU," IEEE International Electron Devices
Meeting (IEDM) , pp. 10.5.1–10.5.4, 2012.
[16] H. Yoda, S. Fujita, N. Shimomura, E. Kitagawa, K. Abe, K. Nomura, H. Noguchi, and J. Ito, "Progress of STT-MRAM Technology and the Effect on Normally-Off Computing Systems," IEEE International
Electron Devices Meeting (IEDM), pp. 11.3.1–11.3.4, 2012.
[17] E. Kitagawa, S. Fujita, K. Nomura, H. Noguchi, K. Abe, K. Ikegami, T. Daibou, Y. Kato, C. Kamata, S. Kashiwada, N. Shimomura, J. Ito, and H. Yoda, "Impact of Ultra Low Power and Fast Write Operation
of Advanced Perpendicular MTJ on Power Reduction for High-Performance Mobile CPU," IEEE
International Electron Devices Meeting (IEDM), pp. 29.4.1–29.4.4, 2012. [18] Everspin Technologies, [Online], Available: http://www.everspin.com/spinTorqueMRAM.php
[19] J. Slaughter, S. Aggarwal, S. Alam, T. Andre, H. J. Chia, M. DeHerrera, S. Deshpande, D. Houssameddine,
J. Janesky, F. B. Mancoff, K. Nagel, M. L. Schneider, N. D. Rizzo, and R. Whig, "Properties of CMOS-Integrated Magnetic Tunnel Junction Arrays for Spin-Torque Magnetoresistive Random Access Memory,"
Abstracts of the 58th Annual Conference on Magnetism and Magnetic Materials, Symposium on Materials
Advances of Spin-Torque Switched Memory Devices for Silicon Integration, BA-01, 2013. [20] D. Worledge, S. L. Brown, W. Chen, J. Harms, G. Hu, R. Kilaru, W. Kula, G. Lauer, L. Q. Liu,
J. Nowak, S. Parkin, A. Pushp, S. Murthy, R. P. Robertazzi, G. Sandhu, and J.Z. Sun, "Materials
Advances in Perpendicularly Magnetized MgO-Tunnel Junctions for STT-MRAM," Abstracts of the 58th Annual Conference on Magnetism and Magnetic Materials, Symposium on Materials Advances of Spin-
Torque Switched Memory Devices for Silicon Integration, BA-02, 2013.
[21] L. Thomas, G. Jan, J. Zhu, H. Liu, Y. Lee, S. Le, R. Tong, K. Pi, Y.Wang, T. Zhong, T. Torng, and P.Wang, "Magnetization Dynamics in Perpendicular STT-MRAM Cells with High Spin-Torque
Efficiency and Thermal Stability," Abstracts of the 58th Annual Conference on Magnetism and Magnetic
238 T. WINDBACHER, H. MAHMOUDI, A. MAKAROV, V. SVERDLOV, S. SELBERHERR
Materials, Symposium on Materials Advances of Spin-Torque Switched Memory Devices for Silicon
Integration, BA-03, 2013. [22] U. Tietze and C. Schenk, Electronic Circuits – Handbook for Design and Applications. 2nd ed. Springer,
2008, no. 12.
[23] W. Zhao, L. Torres, Y. Guillemenet, L. V. Cargnini, Y. Lakys, J.-O. Klein, D. Ravelosona, G. Sassatelli, and C. Chappert, "Design of MRAM Based Logic Circuits and its Applications," In ACM Great Lakes
Symposium on VLSI, pp. 431–436, 2011.
[24] T. Windbacher, H. Mahmoudi, V. Sverdlov, and S. Selberherr, "Spin Torque Magnetic Integrated Circuit," EP 13161375, filed at 27th of March 2013.
[25] R. Hertel, Handbook of Magnetism and Advanced Magnetic Materials. John Wiley & Sons, Ltd, 2007,
Chapter: Guided Spin Waves. [26] M. Donahue and D. Porter, "OOMMF User's Guide," Interagency Report NISTIR 6376, version 1.0.
[27] T. Windbacher, H. Mahmoudi, V. Sverdlov, and S. Selberherr, "Rigorous Simulation Study of a Novel
Non-Volatile Magnetic Flip Flop," In Proceedings of the 18th International Conference on Simulation of Semiconductor Processes and Devices (SISPAD), pp. 368–371, 2013.
[28] T. Gilbert, "A Lagrangian Formulation of the Gyromagnetic Equation of the Magnetization Field," Phys.
Rev., vol. 100, pp. 1243, 1955. [29] H. Kronmüller, Handbook of Magnetism and Advanced Magnetic Materials. John Wiley & Sons, Ltd,
2007, Chapter: General Micromagnetic Theory.
[30] J. Xiao, A. Zangwill, and M. D. Stiles, "Boltzmann Test of Slonczewski's Theory of Spin-Transfer Torque," Phys. Rev. B, vol. 70, pp. 172405, 2004.
[31] J. E. Miltat and M. J. Donahue, Handbook of Magnetism and Advanced Magnetic Materials. John Wiley
& Sons, Ltd, 2007, Chapter: Numerical Micromagnetics: Finite Difference Methods. [32] J. A. Osborn, "Demagnetization Factors of the General Ellipsoid," Phys. Rev., vol. 67, no. 11,12,
pp. 351–357, 1945.
[33] T. Windbacher, H. Mahmoudi, V. Sverdlov, and S. Selberherr, "Novel MTJ-Based Shift Register for Non-Volatile Logic Applications," In Proceedings of the 2013 IEEE/ACM International Symposium on
Nanoscale Architectures (NANOARCH), pp. 36–37, 2013.
[34] M. Tsoi, A. G. M. Jansen, J. Bass, W.-C. Chiang, M. Seck, V. Tsoi, and P. Wyder, "Excitation of a Magnetic Multilayer by an Electric Current," Phys. Rev. Lett., vol. 80, no. 19, pp. 4281– 4284, 1998.
[35] S. I. Kiselev, J. C. Sankey, I. N. Krivorotov, N. C. Emley, R. J. Schoelkopf, R. A. Buhrman,
and D. C. Ralph, "Microwave Oscillations of a Nanomagnet Driven by a Spin-Polarized Current," Nature, vol. 425, pp. 380–382, 2003.
[36] C. Boone, J. A. Katine, J. R. Childress, J. Zhu, X. Cheng, and I. N. Krivorotov, "Experimental Test of an
Analytical Theory of Spin-Torque-Oscillator Dynamics," Phys. Rev. B, vol. 79, pp. 140404-1–140404-4, 2009. [37] G. Finocchio, N. Maugeri, L. Torres, and B. Azzerboni, "Domain Wall Dynamics Driven by a Localized
Injection of a Spin-Polarized Current," IEEE Trans. on Magnetics, vol. 46, pp. 1523-1–1523-4, 2010.
[38] D. V. Berkov, C. T. Boone, and I. N. Krivorotov, "Micromagnetic Simulations of Magnetization Dynamics in a Nanowire Induced by a Spin-Polarized Current Injected via a Point Contact," Phys. Rev. B, vol. 83,
pp. 054420-1–054420-10, 2011. [39] Z. Zeng, G. Finocchio, B. Zhang, P. K. Amiri, J. A. Katine, I. N. Krivorotov, Y. Huai, J. Langer,
B. Azzerboni, K. L. Wang, and H. Jiang, "Ultralow-Current-Density and Bias-Field-Free Spin-Transfer
Nano-Oscillator," Sci. Rep., vol. 3, pp. 1426-1–1426-5, 2013. [40] S. E. Russek, W. H. Rippard, T. Cecil, and R. Heindl. Handbook of Nanophysics: Functional
Nanomaterials, CRC Press, 2010, Chapter: Spin-Transfer Nano-Oscillator, pp. 38-1–38-23.
[41] T. Windbacher, A. Makarov, H. Mahmoudi, V. Sverdlov, and S. Selberherr, "Novel Bias-Field-Free Spin Transfer Oscillator," In Proceedings of the 58th Annual Conference on Magnetism and Magnetic
Materials (MMM), J. Appl. Phys., in press.
[42] W.-G. Wang, M. Li, S. Hageman, and C. L. Chien, "Electric-Field-Assisted Switching in Magnetic Tunnel Junctions," Nat. Mater., vol. 11, iss. 1, pp. 64–68, 2012.
[43] Y. Shiota, T. Nozaki, F. Bonell, S. Murakami, T. Shinjo, and Y. Suzuki, "Induction of Coherent
Magnetization Switching in a Few Atomic Layers of CoFe Using Voltage Pulses," Nat. Mater., vol. 11, iss. 1, pp. 39–43, 2012.
[44] X. He, Y. Wang, N. Wu, A. N. Caruso, E. Vescovo, K. D. Belashchenko, P. A. Dowben, and Ch. Binek,
"Robust Isothermal Electric Control of Exchange Bias at Room Temperature," Nat. Mater., vol. 9, iss. 7, pp. 579–584, 2010.
[45] T. Nozaki, Y. Shiota, S. Miwa, S. Murakami, F. Bonell, S. Ishibashi, H. Kubota, K. Yakushiji, T. Saruya,
A. Fukushima, S. Yuasa, T. Shinjo, and Y. Suzuki, "Electric-Field-Induced Ferromagnetic Resonance Excitation in an Ultrathin Ferromagnetic Metal Layer," Nat. Phys., vol. 8, iss. 6, pp. 491–496, 2012.